Object based video segmentation using multiple cues

In this paper we present a new approach to extract video objects from video sequence by combining of color, motion information and orientation information measure. The motion regions are achieved using a quad-tree split-merge strategy on the motion vector map which is obtained by the different image of several frames in image sequence. The key idea of this approach is to use the orientation information measure and color information to find regions of homogenous characteristics and then use motion difference of image sequences to extract the moving objects. The algorithm use the function of similar region determinant to combine the regions having the similarity spatial and motion information. In order to make better use of available computer resources to achieve optimal performance with a new multi-resolution technique, in which the calculation in the high resolution will only be performed in the boundary of the sub-region, to significantly accelerate the computation. Finally, To obtain the accurate region boundaries, the information of color image edge detection is used.